摘要
针对高校监考任务繁重且传统监考作弊取证较难等问题,设计了基于视频行为分析的智能监考辅助系统。该系统首先应用视频流实时处理算法对视频流进行解码,然后运用YOLOv3算法检测出人体以及违禁品边框,接着用MTCNN算法检测面部姿态,最后对行为异常进行检测分析并给出异常警告。通过使用该系统,能够实现自动检测考生作弊行为并且可以自动预警和保存作弊行为。经过测试,系统的识别结果准确率能够达到百分之五十以上,能够起到辅助监考的作用。
In view of the heavy task of invigilating in colleges and universities and the difficulty of cheating and obtaining evidence in traditional invigilation, an intelligent invigilating assistant system based on video behavior analysis is designed. The system first uses the video stream real-time processing algorithm to decode the video stream, then uses the YOLOv3 algorithm to detect the human body and the contraband frame, and then uses the MTCNN algorithm to detect the facial posture. Finally, the behavior anomaly is detected and analyzed and the anomaly warning is given. By using the system, the cheating behavior of candidates can be detected automatically and the cheating behavior can be automatically warned and saved. After testing, the recognition accuracy of the system can reach more than 50%, which can play an auxiliary role in invigilating.
作者
李春梅
邵新慧
刘玲
LI Chunmei;SHAO Xinhui;LIU Ling
出处
《科技创新与应用》
2019年第18期8-10,共3页
Technology Innovation and Application
